- Generated by
1.9.1
|
ZigZag - Deep Learning Hardware Design Space Exploration
This repository presents the novel version of our tried-and-tested hardware Architecture-Mapping Design Space Exploration (DSE) Framework for Deep Learning (DL) accelerators. ZigZag bridges the gap between algorithmic DL decisions and their acceleration cost on specialized accelerators through a fast and accurate hardware cost estimation.
|
Namespaces | |
| zigzag.visualization.results.plot_cme | |
Functions | |
| def | shorten_onnx_layer_name (str name) |
| Names generated in the ONNX format are quite long (e.g. More... | |
| def | get_mem_energy_single_op (CostModelEvaluation cme, LayerOperand op, int mem_level) |
| def | get_energy_array (list[CostModelEvaluation] cmes, list[LayerOperand] all_ops, list[MemoryInstance] all_mems) |
| Convert the given list of cmes to a numpy array with the energy per layer, memory level, operand and data direction. More... | |
| def | get_latency_array (list[CostModelEvaluation] cmes) |
| def | bar_plot_cost_model_evaluations_breakdown (list[CostModelEvaluationABC] cmes, str save_path) |
Variables | |
| int | SMALLEST_SIZE = 10 |
| int | SMALLER_SIZE = 12 |
| int | SMALL_SIZE = 14 |
| int | MEDIUM_SIZE = 16 |
| int | BIG_SIZE = 18 |
| int | BIGGER_SIZE = 20 |
| size | |
| titlesize | |
| labelsize | |
| fontsize | |
| int | BAR_WIDTH = 1 |
| int | BAR_SPACING = 0 |
| int | GROUP_SPACING = 1 |
1.9.1